What UCP and ACP mean for retail (and retail media)

Agentic commerce is no longer a theoretical edge case. AI systems are actively moving from answering questions to making decisions and completing transactions. As that shift accelerates, the industry is converging around new standards that redefine how discovery, decisioning, and payments work.

Two protocols sit at the center of this shift: the Universal Commerce Protocol (UCP) and the Agentic Commerce Protocol (ACP).

They are driven by different players and start from different assumptions. While they overlap in meaningful ways, they do not solve the same problems: UCP requires retailers to take action on their data structures and interaction points, while ACP requires them to take action on how they interact with agents.

Together, they signal a deeper structural change in how commerce operates and how retail media must evolve. In this article, we unpack what these two different protocols mean and how retailers need to respond today.

Universal Commerce Protocol (UCP): Making commerce executable everywhere

The Universal Commerce Protocol, driven by Google and a consortium of industry partners, is fundamentally about interoperability.

UCP defines a standardized way for retailers to expose data and operational processes (or skills) via a common language so they can be accessed programmatically by AI agents across many environments. Its goal is to eliminate bespoke integrations and create a shared language for execution.

UCP spans the full commerce lifecycle by exposing the signals AI agents need to execute a purchase without human intervention, including:

• Product and catalog data
• Pricing and promotions
• Inventory availability
• Fulfillment and delivery options
• Checkout and transaction endpoints

UCP does not attempt to influence consumer choice. It creates a common language for AI agents to interact with various merchant systems so they can answer the question: Can this product be bought right now, and how?

As discovery becomes more distributed across agents, UCP ensures retailers’ product data is reachable. For now, retailers don’t need to worry about making their products stand out; they just need to provide the structured data so agents will find the right, relevant product. 

UCP ensures you can be considered. It does not ensure you are chosen.

Agentic Commerce Protocol (ACP): Making buying native to AI

The Agentic Commerce Protocol, driven by OpenAI and Stripe, approaches commerce from the opposite direction.

ACP is centered on agent-led buying experiences, where AI systems interpret intent, evaluate options, and execute transactions on behalf of consumers, often entirely inside conversational or agent-native environments. 

ACP focuses on:

• Agent reasoning and prioritization
• Conversational discovery and evaluation
• Secure authorization and payment execution
• Transaction orchestration and merchant handoff

Rather than standardizing the entire commerce surface, ACP optimizes for the buying experience. Buying becomes part of the interaction, not a destination users are sent to.

ACP answers a different question: Should this be the product the consumer buys right now?

Overlap without convergence?

It’s tempting to describe UCP and ACP as clean layers. In practice, the boundaries are messier.

There is real overlap, especially around execution and payments. Both visions aim to abstract checkout, enable trusted agents to authorize purchases, and support multiple wallets and credentials without requiring users to navigate traditional, page-based checkout flows.

This is where UCP’s trajectory toward agent-to-payment (A2P or Agent Payments Protocol) overlaps with ACP’s built-in payment capabilities. The end state looks similar in that payments become embedded inside agent execution.

The difference is emphasis: UCP starts from commerce systems and works outward, while ACP starts from the agent experience and works inward.

There are also important gaps in ACP’s scope that matter for scaled retailers. ACP does not fully address:

• Open discovery across many environments, where products need to be eligible and accessible beyond a single agent or interface
• Broad catalog exposure and eligibility frameworks, which large retailers rely on to govern what can be sold, when, and under what conditions
• Traditional or hybrid checkout experiences, which will continue to coexist alongside agent-led flows
• Comprehensive commerce capability discovery, required for retailers managing complex catalogs, fulfillment models, and regional constraints

These areas remain squarely within UCP’s domain, reinforcing its role as foundational infrastructure rather than a competing experience layer.

A2A and A2P: Interaction patterns inside the shift

As agent-led buying becomes more common, the mechanics of how agents coordinate and transact start to matter as much as the protocols themselves. Two interaction patterns are emerging as foundational.

A2A (agent-to-agent) describes how agents coordinate across the ecosystem. A consumer-facing agent may interact with retailer agents, inventory systems, or fulfillment services to validate availability, negotiate substitutions, or apply loyalty logic.

A2P (agent-to-payments) describes how agents execute payment actions within ACP flows. Authorization, wallet selection, credential use, routing, and settlement happen programmatically, governed by trust and permissions rather than pages.

ACP enables payments, but A2A and A2P explain how agent-led coordination and payment execution scale across retailers, partners, and environments.

The risk of divergence: A protocol cold war?

While the industry pushes for interoperability, we must also acknowledge a realistic friction: UCP and ACP represent a potential philosophical fork in how the internet is monetized.

At the core, this is fundamentally about who owns the buying relationship.

Google’s UCP vision is fundamentally merchant-first. It preserves a world where retailers expose their capabilities, discovery remains distributed, and monetization still flows through an open ecosystem that search, ads, and comparison can operate on. That aligns cleanly with Google’s long-term interest in keeping the web crawlable and monetizable at scale.

OpenAI and Stripe’s ACP vision is agent-first. It prioritizes completion over exploration, making the user’s primary relationship with the agent. That model favors centralized execution, embedded payments, and fewer visible seams.

This tension creates three specific risks that retailers must weigh against the “coexistence” narrative:

1. The “double-stack” tax: If these protocols do not converge, retailers face the burden of maintaining a “double stack”: a complex data infrastructure for Google’s crawling/API needs (UCP) and a separate set of secure, tokenized payment rails for OpenAI’s ecosystem (ACP). For mid-sized retailers, the resource cost of maintaining two distinct commerce layers may force a choice, leading to fragmentation where brands are only “available” in certain AI ecosystems.
2. The walled garden “opt-out”: The biggest variable is not Google or OpenAI, but Amazon. With its Rufus agent and massive closed marketplace, Amazon has little incentive to empower a neutral “universal” protocol that levels the playing field for independent merchants. If Amazon ignores both UCP and ACP in favor of its own proprietary agent interface, the commerce landscape fractures into three: UCP, ACP, and whatever Amazon does.
3. Distribution scale vs. conversational depth: The battle will likely be fought on two fronts. Google bets that distribution wins and that retailers will adopt UCP because they cannot afford to disappear from Search. OpenAI bets that experience wins and that high-value consumers will migrate to agents that can “just do it” via ACP, leaving friction-heavy search results behind.

Why this fundamentally changes retail media 

Retail media was built for human browsing behavior, where sponsored placements assume that visibility drives choice. Agentic commerce breaks that assumption.

In agent-driven environments, influence happens before visibility. As a result, monetization moves away from selling placement and toward influencing how agents evaluate and select products, including:

• Eligibility and relevance
• Performance and outcome history
• Availability and fulfillment confidence
• Value, loyalty, and trust signals

Retail media does not disappear. It matures. The opportunity moves from selling shelf space to shaping decisions.

What retailers should do next

Retailers do not need to choose between UCP and ACP, but they do need to act deliberately.

1. Treat UCP as a foundation. UCP is infrastructure. It ensures participation. Retailers that delay risk quiet exclusion from agent-driven discovery.
2. Understand where decisions are moving. ACP shifts influence to the moment of decision. Retailers must assert control through the signals and rules that shape agent decisions rather than relying on page design to influence outcomes.
3. Prepare for agent-mediated payments. Checkout will become invisible in many flows, and payment authorization will be delegated to trusted agents. Loyalty, wallets, and credentials will become programmable inputs.
4. Watch incentives, not announcements. Google is optimizing for scale and interoperability while OpenAI is optimizing for experience and intelligence. Payment networks will follow trust and authorization flows.

Retailers remain central, but their leverage now depends on how clearly they express intent, value, and constraints in machine-readable form.

Koddi’s perspective

From Koddi’s point of view, these shifts do not disintermediate retailers. They do, however, redefine retailers’ role in discovery and purchasing cycles. Retailers should look at these new protocols through the lens of opportunity: now, they are able to influence consumer decisions within agentic experiences if they can align with these protocols. 

Overall, UCP makes commerce accessible, and ACP makes commerce autonomous. As commerce signals become standardized and agents take on more responsibility, advantage moves to systems that intentionally shape outcomes. 

In 2026, Koddi will help retailers turn inventory, availability, performance, and value into measurable influence across environments, whether execution happens directly onsite, through UCP-driven surfaces and/or ACP-driven agents. That is where the future of retail and retail media will be won.